Hi! I am Jinyu Xu (å¾čæå®), an undergraduate student at the School of Cyber Science and Engineering, Huazhong University of Science and Technology (HUST). My research interests include Computer Graphics and Computer Vision.
I am interested in what limited visual observations can tell us about the world, from the appearance captured in images to the hidden order behind it. Advised by Prof. Hao Lu, I have worked on visual dense prediction, studying pixel-level visual understanding in tasks such as dense object counting and fine-grained plant phenotyping.
Please feel free to reach out via jinyu_xu@hust.edu.cn for discussion or collaboration.
š Publications
Plant Taxonomy Meets Plant Counting: A Fine-Grained, Taxonomic Dataset for Counting Hundreds of Plant Species
Jinyu Xu, Tianqi Hu, Xiaonan Hu, Letian Zhou, Songliang Cao, Meng Zhang, Hao Lu
CVPR 2026 Oral (Best Paper Award Candidate)
Paper | Code | Dataset | Project Page | Social Media
This work introduces a large-scale, fine-grained plant counting dataset spanning hundreds of species. By integrating botanical taxonomy, we aim to bridge the gap between plant science and computer vision to enhance cross-species generalization.
TasselNetV4: A vision foundation model for cross-scene, cross-scale, and cross-species plant counting
Xiaonan Hua, Xuebing Li, Jinyu Xu, Abdulkadir Duran Adan, Letian Zhou, Xuhui Zhu, Yanan Li, Wei Guo, Shouyang Liu, Wenzhong Liu, Hao Lu
ISPRS Journal of Photogrammetry and Remote Sensing 2026
Paper | Code | Model | Social Media
This work defines Plant Agnostic Counting, a task focusing on zero-shot generalization across various species. We provide two datasets, PAC-105 and PAC-Somalia, and propose TasselNetV4 as a robust model for precision plant counting.
Revisiting unsupervised image stitching via efficient boundary rectification
Yun Zhang, Jialing Yang, Ruiyang Liang, Yao Xu, Lang Nie, Jinyu Xu, Fang-Lue Zhang, Xinyuan Zheng
Computers & Graphics 2026
Paper | Code
BRecStitch introduces the first end-to-end unsupervised framework for image stitching and rectification, balancing alignment and regularity via bidirectional warping and a differentiable boundary loss.
Enhancing trustworthiness evaluation of Large Language Models through dataset refinement
Jinwei Xu, Xin Zhou, Yanjing Yang, Tianhao Liu, Jinyu Xu, Chang Zhang, Lanxin Yang, Liming Dong
Information and Software Technology 2026
Paper
This work proposes TrustDA, a dataset refinement framework that improves LLM trustworthiness evaluation by reducing evaluation costs and mitigating inflated benchmark scores.
š Honors and Awards
- 2025.10, National Scholarship (Highest honor for undergraduates in China).
- 2025.10, Merit Student, Huazhong University of Science and Technology.
- 2025.05, Excellent Undergraduate Student, Huazhong University of Science and Technology.
- 2024.10, National Scholarship (Highest honor for undergraduates in China).
- 2024.10, Merit Student, Huazhong University of Science and Technology.
š¬ Invited Activities
- 2026.5, Poster Presenter, Vision And Learning SEminar (VALSE).
- 2026.5, Poster Presenter, China Society of Image and Graphics (CSIG) Conference.
š» Internships
- Research Intern, State Key Laboratory for Novel Software Technology, Nanjing University.
- Worked on the compliance assessment of multi-party large language models (LLMs).
š Educations
- 2023.9 - 2027.6 (expected), B.Eng, Information security, Huazhong University of Science and Technology.
šØ Hobbies
Beyond my research, I enjoy exploring different forms of art:
- Calligraphy: I enjoy practicing calligraphy to stay focused and appreciate the beauty of its structure and lines.
- Movies: Iām a big movie fan who loves exploring different stories and occasionally sharing my thoughts in reviews.